doosra

Justin, XLA sounds interesting. Do you assume you always have CUDA sources for ML operations in XLA? I was under the impression that closed-source libraries like cuDNN were used.

Is it possible to accurately evaluate the profitability of fusing two kernels in CUDA (effects of increased register pressure; shared memory)? On the other hand, the generic kernel and its launch parameters were probably hand tuned for performance.

Does anyone know what kind of sensor can detect levels of pollen in a personal space? Are they available for purchase at a reasonable price?

I understand your frustration. I too graduated from such a private college with a CS degree in 2003. Here is why I think you should complete your degree:

- You are being forced to study the core areas in CS. You will probably soon specialize in your work/research, so learn the core areas when you still have the time.

- Unless you have already created exceptional products, regular employers will not consider you unless you have a degree; these employers are a good fallback if nothing else works. The degree is a baseline to avoid being filtered out.

- If you want to go to graduate school to do research, either now or in the future, an undergraduate degree is essential.

- Getting a B.Tech isn't hard. Put in the minimum effort if you want, but it isn't tough to pass, nor is it a time sink.

Yes, the teaching can be unimaginative, professors may strictly adhere to a syllabus, and the exams test rote memorization, but one can still be stimulated in this environment. I would unceasingly ask questions to fully understand algorithms being taught in class, often challenging professors (and finding holes in the thought process). The database class helped me understand normalization and other ideas of good DB design (and their rationale), which I used for my web projects. A queuing theory class helped me understand performance of a system I later built.

There is no doubt you need to go beyond just getting the degree. Do you have fantastic ideas for products? Maybe a bit of freelancing will help? I created websites for clients abroad while doing my undergrad (just get a low bandwidth internet connection from a mobile company for this). I interned with companies/organizations during the summer. For my final year project I wrote a simd+cluster version of a program that I ran on all the computers in the department's lab (and tripped the ups!). I also realized the importance of a degree and tried to stay close to the top of my graduating class.

The challenge is to stay motivated to get the degree, while also finding an outlet for your passion. Both can be done. Freelance during the semester working nights and weekends; work with a company during the summer; participate in CS competitions in your region; find a local conference in an area of interest and publish a paper (this isn't hard!). It's really up to you to get motivated and rise above the rest.

While I sometimes wish I had done my undergrad elsewhere, I was eventually able to work in an area of my interest. I went to graduate school abroad and am now working in a research department in industry doing what I love.

I was hoping to see a speed comparison of the two implementations.

The way I think Intel can get competitive is by building a system on a chip. Think a cpu, gpu, and perhaps other cores on a single chip, consuming relatively less power. Nvidia is doing that with an arm cpu.

Does any one know the specs of the FPGA? How many LUTS/block rams?

I think its exciting that Intel has started to pay attention to FPGAs, and hopefully with mass production, this FPGA+CPU chip will be priced cheaply---that may yet be the biggest contribution here.

A Bloom filter (http://en.wikipedia.org/wiki/Bloom_filter) might work well in this case.

I love this feature, do you know if its possible to do that without logging into their website? I would love to be able to create temp numbers from my phone (without having to log in and expose the rest of the functionality).

How does this array do I/O? How does one bring in data (from disk/network) into the chip? I couldn't find anything demonstrating high throughput I/O on the eval board...

I love the idea of Plunify. I tried to find pricing information (for future use) on the website, to no avail. But if this cuts down the cost for CAD tools as an independent developer, it will be huge for me!

I started off hating Python as well. I could not stand a language that didn't terminate statements without semicolons! :) And python just seemed a lot like Perl; when using both languages for large projects my code just seemed to degenerate to "hacky."

But I've grown to like the simplicity of Python. One thing I don't like is its threads implementation, however. Because of the GIL, multithreading (GUI + background thread) just doesn't work right. Maybe I'm using it wrong?

The fine-grained control is a plus for openDNS.

Second, from the stories that seem to pop up, google's achilles heel is it's customer service. OpenDNS might be able to do better.

Ahh... the author of the gamesbrief article seems to be talking about the situation in the UK. It seems a lot different in the US. I wonder if the gender discrimination the author talks about happening in the UK also happens in the US despite considerably laxer laws.

I googled and found something online: http://www.forbes.com/2009/05/04/maternity-leave-laws-forbes...

According to the article the law grants 12 weeks unpaid time off to care for a newborn baby. The old job is guaranteed on the mother's return. The woman must also have worked at the company for at least one year. Moreover, companies with less than 50 employees are exempt from this law so most startups are not bound.

I don't find this law unreasonable.

> But most people seriously just need a way to check email and login to Facebook

But why not just use a cell phone? Apart from the large screen, it seems to me that a chrome-os system will be similar to a smartphone.

Wait, just by chance it looks like there was a 4/5 chance that the Stradivarius would loose/be outdone.

The big problem I see in the implementation of this idea: How does one forum handle the mass of users? I'm not talking about technical issues, more of the user experience. How can all users engage in the discussion without being overwhelmed?

Help me understand the business model of GM crops: Does Monsanto create GM crops that are seedless? I think this is evil since it gives too much power to one corporation and removes the farmer's prerogative. On the other hand, how will Monsanto make money after the first crop cycle?

"Google... said users would no longer have to worry about viruses, malware and security updates"

I'm guessing since most applications on Chrome will be web-based, the vendor will worry about them. They will be able to more easily and quickly detect + destroy phishing schemes, viruses etc. Kind of like how Facebook has responded to malicious wall posts.

For me the bigger concern would be the loss of productivity due to downtime of web services, or loss of internet connectivity.

There are pros/cons for govt. subsidies. Some third world countries subsidize seeds/fertilizers/electricity for poor farmers. Many times this is barely enough for subsistence for them and their large families. I'm not sure what a feasible alternative is.

Not all nations go to war of course. But an imbalanced trade equation, especially for basic necessities, does at the very least lead to a loss of leverage at the negotiation table.

I'm not saying subsidies are good; all I'm saying is I can understand why the US govt. would want to maintain a source of homegrown food supply.

The purpose of subsidies is food security for the US. Importing food at a cheaper rate will likely adversely impact local production, which may be important in the future. The US is already dependent on other countries for its energy supplies. Nations go to war over such stuff.

While only speaking from experience, I've noticed that many foreign PhDs in CS, Electrical and other engineering fields are increasingly electing to go back once they are done.

The main reasons seem to include being so far away from family and being in an alien culture. Further, jobs for PhDs are increasingly available in their home countries-- this seems to seal the deal. The tough immigration stance contributes to the discouragement since foreign workers have to "pause" their lives until they get their green card.

I find this to be true with news reporting as well. Compare CNN v. BBC.

The problem feature "allows users to quickly find out whether people they know are on Facebook and invite them to become a friend by uploading a .txt or .csv file of email addresses."

Facebook would reveal alternate email addresses of a user if at least one matched the .txt list. This bug was fixed.

The feature can still be used to validate email addresses, for example by spammers. Further, many social networks display limited information about the user, which can be useful demographic information to a spammer, if the email address is known (guessed correctly).

This feature is very useful and is critical in quickly expanding one's social network. Unfortunately, it seems it can be easily exploited. How would one prevent misuse? Limiting the number of email addresses checked in the .txt file may be one partial solution.

scp -C does the same thing. Both methods use gzip for compression.

Does anyone know of a similar reader at about the same price point which supports annotation of the text (for example with a stylus)?

Fascinating thoughts, though most of the points seem to be justified using anecdotes. I think many of these points should be encouraging to all entrepreneurs.

I also think the tech industry is missing out by not having more women participate.

The author says "instead of scouring the Web for interesting stories, everything came to me!"

I believe the author is confusing the presentation of content with the filtering. The author went from using a browser, to an RSS reader, back to a browser. What remains unsolved is the problem of filtering uninteresting stories.

The real reason his new solution works is because he's manually filtered stories. But this approach doesn't scale with the number of websites that are being tracked.

What remains unsolved is an algorithm to find stories that are interesting to the author.